MATH 106 LINEAR ALGEBRA

LECTURE NOTES
FALL 2010-2011

These Lecture Notes are not in a final form being still subject of improvement

Contents
1 Systems of linear equations and matrices

1.1

Introduction to systems of linear equations . . . . . . . . . . . . . . . . . . . . . . .

1.2

Gaussian elimination . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

1.3

Matrices and matrix operations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12

1.4

Inverses. Rules of matrix arithmetic . . . . . . . . . . . . . . . . . . . . . . . . . . . 18

1.5

Elementary matrices and a method for finding A1 . . . . . . . . . . . . . . . . . . . 22

1.6

Further results on systems of equations and invertibility . . . . . . . . . . . . . . . . 25

1.7

Diagonal, Triangular and Symmetric Matrices . . . . . . . . . . . . . . . . . . . . . . 27

2 Determinants

33

2.1

The determinant function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33

2.2

Evaluating determinants by row reduction . . . . . . . . . . . . . . . . . . . . . . . . 35

2.3

Properties of determinant function . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36

2.4

Cofactor expansion. Cramers rule . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38

3 Euclidean vector spaces

43

3.1

Euclidean n-space . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43

3.2

Linear transformations from Rn to Rm . . . . . . . . . . . . . . . . . . . . . . . . . . 46

3.3

Properties of linear transformations from Rn to Rm . . . . . . . . . . . . . . . . . . . 50

Chapter 1

Systems of linear equations and

matrices
1.1

Introduction to systems of linear equations

Definition 1.1.1 A linear equation in the n variables x1 , . . . , xn , equally called unknowns, is the
problem of finding the values of x1 , . . . , xn such that a1 x1 + . . . + an xn = b, where a1 , . . . , an , b are
constants. A solution of a linear equation a1 x1 + . . . + an xn = b is a sequence t1 , . . . , tn of n real
numbers such that the equation is satisfied when we replace x1 = t1 , . . . , xn = tn . The set of all
solutions of the equation is called its solution set or the general solution.

Definition 1.1.2 A finite set of linear equations in the variables x1 , . . . , xn is called a system of
linear equations or a a linear system. A solution of a linear system

a11 x1 + a12 x2 + . . . + a1n xn = b1

a21 x1 + a22 x2 + . . . + a2n xn
..
.

= b2
..
.

(1.1)

am1 x1 + am2 x2 + . . . + amn xn = bm

with m linear equations and n unknowns, is a sequence t1 , . . . , tn of n real numbers such that each
equation of the system is satisfied when we replace x1 = t1 , . . . , xn = tn . The set of all solutions of
the linear system is called its solution set or the general solution. A system of equations is said to
be consistent if it has a solution at least. Otherwise it is called inconsistent. The augmented matrix

of the linear system (1.3) is

a11

a21

..
.

am1

a12

...

a1n

b1

a22
..
.

...
..
.

a2n
..
.

b2
..
.

am2

. . . amn

bm

Remark 1.1.3 The solution set of a linear system unchanges if we perform on the system one of
the following operations:
1. Multiply an equation with a nonzero constant;
2. Interchange two equations;
3. Add a multiple of one equation to another.
The corresponding operations at the level of augmented matrix are:
1. Multiply a row with a nonzero constant;
2. Interchange two rows;
3. Add a multiple of one row to another.

Example 1.1.4 Solve the following linear system

x + 2y 3z +

aw =

3x

y + 5z + 10aw = 2 .

4x +

y + 2z + 11aw =

Solution: The augmented matrix of the system is 3 1

4
1

4
a

5 10a 2 and we have

2
2 11a

2 3

successively:

4
1
2 3 a

3 1
5 10a 2

4
1
2 11a
2

x + 2y 3z +

aw =

3x

y + 5z + 10aw = 2

4x +

y + 2z + 11aw =

3r1 + r2 r2 4r1 + r3 r3

3e1 + e2 e2 4e1 + e3 e3

4
1
2 3 a

0 7 14 7a 14

0 7 14 7a 14

x + 2y

e2 + e3 e3

a
4

14 7a 14

0 0
0

2 3

0 7

0
0

x + 2y

17 e2

1 2 3 a 4

0 1 2 a 2

0 0
0 0
0

x + 2y 3z + aw = 4
y 2z aw = 2
0 = 0

2r2 + r1 r1

2e2 + e1 e1

1 0
1 3a 0

0 1 2 a 2

0 0
0
0 0
Thus, the equivalent system is

aw =

0 =

3z +

7y + 14z + 7aw = 14

17 r2

7y + 14z + 7aw = 14

aw =

7y + 14z + 7aw = 14

r2 + r3 r3

3z +

z + 3aw =

y 2z

aw =

0
2

0 = 0.
+z +3aw = 0

y 2z

aw

and its solutions are

=2

x = t 3as, y = 2 + 2t + as, z = t, w = s for s, t R.

1.2

Gaussian elimination

Definition 1.2.1 A matrix is said to be in reduced row-echelon form if it has the following properties:
1. If a row does not consist entirely of zeros, then the first nonzero number in the row is a 1,
called a leading 1.

2. If there are any rows that consist entirely of zeros, then they are grouped together at the
bottom of the matrix.
3. In any two successive rows that do not consist entirely of zeros, the leading 1 in the lower
row occurs farther to the right than the leading 1 in the higher row.
4. Each column that contains a leading 1 has zero everywhere else.
A matrix satisfying just the first 3 properties, namely 1,2 and 3, is said to be in row-echelon form
We are now going to provide a number of steps in order to reduce a matrix to a (reduced) rowechelon form. These are:
1. Locate the lefmost column that does not consist entirely of zeros;
2. Interchange the top row with another row, if necessary, to bring a nonzero entry to the top
of the column found in Step 1;
3. If the entry that is now at the top of the column found in Step 1 is a, multiply the first row
by

1
a

in order to introduce a leading 1;

4. Add suitable multiples of the top row to the rows below so that all entries below the leading
1 become zeros;
5. Cover the top row in the matrix and begin again with the Step 1 applied to the submatrix
that remains. Continue in this way until the entire matrix is in row-echelon form.
6. Once you got the matrix in row-echelon form, beginning with the last nonzero row and working
upward, add suitable multiples of each row to the rows above to introduce zeros above the
leading 1s.
The variables corresponding to the leading 1s in row echelon form are called leading variables
and the others are called free variables.
The above procedure for reducing a matrix to reduced row-echelon form is called Gauss-Jordan
elimination. If we use only the first five steps, the procedure produces a row-echelon form and is
called Gaussian elimination.
Examples 1.2.2

1. Solve the linear system

x
2x +

y + 2z

w = 1

y 2z 2w = 2

x + 2y 4z +
3x

w =

3w = 3

(1.2)

and it can be reduced at

1 1
2 1

2
1 2 2
Solution: The augmented matrix is

1
2 4
1

3
0
0 3
reduced row-echelon form as follows:

1 1
2r +r r

2
2

2 2 r1 + r3 r3

3r1 + r r

4
4

1
1

3 3

1 1
2 1

0
1 2
0
r2 + r3 r3

3r2 + r4 r4 0
0
0
0

0
0
0
0

1 1
1

0 3 r2 0
0

0
0
0

0
0
0

0 r2 + r1 r1

1 0

0 1

0 1 2

0 0

0 0

such that the corresponding linear system, equivalent with the initial one, is
x=t1
x

w = 1
y 2z

and its solutions are

y = 2s

, s, t R,

z=s
w=t

being infinitely many in this case. It is sometimes preferable to solve a linear system by
using Gauss elimination procedure for the augmented matrix to bring it just in a row-echelon
form and to use the so called back-substitution method afterward. For the linear system 1.2,
a row-echelon form of the augmented matrix is

1 1
2 1 1

0
1 2
0
0

0
0
0
0
0

0
0
0
0
0
and its corresponding linear system is
x y + 2z w = 1
y 2z

The back-substitution method consists in the following three steps:

(a) Solve the equations for the leading variables in terms of the free variables;
(b) Beginning with the bottom equation and working upward, successively substitute each
equation into all the equations above it.

1 1

0
0

0
0

0
0

(c) Assign arbitrary values to the free variables, if any.

In our particular case x, y are the leading variables and z, w are the free variables. Therefore
the step (1a) is
x = y 2z + w 1
y = 2z
The step (1b) is
x = 2z 2z + w 1

or equivalently

y = 2z

x = w1
y = 2z

Finally, the step (1c) consists in assigning to the free variables z, w the arbitrary values z = s
and w = t such that the solutions of the initial linear system are:
x=t1
y = 2s

, s, t R.

z=s
w=t
2. Solve the linear system
x + 7y 2z = 1
3x

y +

z =4

2x + 6y

z =5

Solution: The augmented matrix is

row-echelon form as follows:

3 1
2

1
2 1

1
3
1 4

2
1 5

7
1
6

7 2 1

3 1
1 4 and it can be reduced at reduced

2
6 1 5

1
2 1

3r + r r

1
2
2
0
1
4

2r1 + r3 r3

0
1
5

2 1
1 7
2

1/20r

2
0
7
20 5
1 41

0
0
0
0
0
0

1 7

7 7r2 + r1 r1
20

7
20
20

2 1

r + r r

2
3
3

5
7

5
7

29

20

1
0 1 4
7/20

0 0
0
0
1 0

such that the corresponding linear system, equivalent with the initial one, is
x

1
4z

29
20

1
4z

= 7/20

and its solutions are

x=

295t
20

y=

7+5t
20

z = t.

10

t R.

1
4

3. Solve the linear system

x1

2x2

x3

4x4

=1

x1

3x2

7x3

2x4

=2

x1

12x2

16x4

=5

11x3

1 2
1 4 1

Solution: The augmented matrix is 1

3
7
2 2 and it can be reduced at

1 12 11 16
5
reduced row-echelon form as follows:

1 2
1 4 1
1 2
1 4 1

r + r r

2r + r r

2
1
2
2
3
3
0

1
3
7
2 2
5
6
6 1

r1 + r3 r3

1 12 11 16 5
0 10 12 12 4

1 2 1 4 1

2r2 + r3 r3

0
5 6
6 1 .

0
0 0
0 6

Thus the system is inconsistent, that is it has no solutions at all, since 0 6= 6.

4. For which values of a R the following linear system has no solution ? Exactly one solution
? Infinitely many solutions ? When the system is consistent, find its solution set.
x

2z =

2x 4y

6z =

3x 5y + (a2 14)z = a 4

Solution: The augmented matrix is 2 4

6 6

2
3 5 a 14 a 4
performing successively the following row-operations:

1 1

1
4
5

8 6

a2 14 a 4
2

21 r2

1
1
2

2r + r r

1
2
2
0

3r1 + r3 r3
0

2
0

2
3

a2 8 a 4

1
2
2

2r + r r
2
3
3

0
|

1
1
0

and it will be reduced by

4 6

a2 8 a 4
2

2
0

2
3

a2 4 a + 2
{z
()

11

1r
2 2

a2 4

if a 6 {2}

0
if a 6 {2}
0
1
a2 4

1
1
0

2r3 +r2 r2

2
0
a2

2
0 3 a2

1
1
a2

2r + r r

1
1
0

1
if a 6 {2}

1
0
0
a2

1 0 0
3

r2 + r1 r1

0 1 0 3a8
.
a2
if a 6 {2}

1
0 0 1 a2
3

Therefore for a 6 {2} the given linear system has the unique solutions x = 3, y =
If a = 2, the () reduced matrix becomes

0
1
1 1 2
r +r r

1
1
0
0
1 2 3

0
0
0 0 0

0
1
0

r +r r
2
1
1

if a 6 {2}

3a8
a2 , z

1
a2 .

3
0

2 3 ,

0 0

and in this case the given linear system has infinitely many solutions
x = 3, y = 3 2s, z = s, s R.

1 1 2

Finally, when a = 2, the () reduced matrix becomes 0

1 2

0
0 0
solutions at all in this case.

1.3

3 and linear system has no

Matrices and matrix operations

Definition 1.3.1 A matrix

A=

a11

a12

a1n

a21
..
.

a22
..
.

a2n
..
..
.
.

am1 am2 amn

row 1
row 2
..
.
row m

col.1 col.2 col.n

equally written [aij ]mn (or simply [aij ]), is a rectangular array of numbers. The numbers in the
array are called entries. The entry in row i and column j of a matrix A is commonly denoted by
(A)ij , in our particular case (A)ij = aij . The size of a matrix is described in terms of the number
of rows and the number of columns it has. The above matrix A has size m n. A matrix

x1

x2

x= .

..

xm

12

with only one column is called a column matrix (or a column vector), and a matrix
h
y=

i
y1

y2

. . . yn

with only one row is called row matrix (or a row vector). A matrix

a11

a21
A=
..
.

an1

a12

. . . a1n

a22
..
.

. . . a2n
..
..
.
.

an2

. . . ann

with n rows and n columns, that is of size n n, is called a square matrix of order n and the entries
a11 , a22 , . . . , ann are said to be on the main diagonal of A.
Definition 1.3.2 Two matrices are said to be equal if they have the same size and their corresponding entries are equal.
Definition 1.3.3 If

a11

a21
A=
..
.

am1

a12

...

a1n

a22
..
.

...
..
.

a2n
..
.

am2

. . . amn

b11

b
, B = 21

..

bm1

b12

...

b1n

b22
..
.

...
..
.

b2n
..
.

bm2

. . . bmn

are matrices of the same size, then their sum is the matrix

a11 + b11

a12 + b12

...

a1n + b1n

+ b21
..
.

a22 + b22
..
.

...
..
.

a2n + b2n
..
.

+ bm1

am2 + bm2

. . . amn + bmn

a11 b11

a12 b12

...

a1n b1n

b21
..
.

a22 b22
..
.

...
..
.

a2n b2n
..
.

bm1

am2 bm2

. . . amn bmn

a21
A+B =

am1

and their difference is the matrix

a21
AB =

am1

One can shortly define the entries of the sum and the difference matrices in the following way:
(A + B)ij = (A)ij + (B)ij and (A B)ij = (A)ij (B)ij .

13

If the matrices A and B have different sizes, then their sum and difference is not defined. If c is
any scalar (number), then the product cA is the matrix

ca11 ca12 . . . ca1n

ca21 ca22 . . . ca2n

..
..
..
..
.
.
.
.

cam1

cam2

. . . camn

obtained by multiplying each entry of A with c and its entries can be shortly written as (cA)ij =
c(A)ij . If A1 , A2 , . . . , An are matrices of the same size and c1 , c2 , . . . , cn are scalars, then the matrix
c1 A1 + c2 A2 + + cn An is defined and it is called a linear combination of A1 , A2 , . . . , An with
coefficients c1 , c2 , . . . , cn .
Examples 1.3.4 Consider the matrices

A=

1 3 0
2

, C =
1 0
4

3
2 5 1

1 2 2

, B =

0 2

2 4

1 2 3

, D =

1 6

3 .

1 2 4
2 1

1. Compute, when is possible, A B, C + D, B C, A + C,

Solution:

AB =

1 3 0

D+E =

1 2 2

0 5

=
2 4
3
4


1 2 3
1 6
2




1 0
4 + 2 1
3 =


2 5 1
1 2 4

0 2

4 1

0 8 1

1 1
7 .

3 7 5

B C and A + C are not defined.

1
D. Solution:
2. Compute, when possible, 13 A, (1)B, 2A B, 3C + D, 5B + 2006

1
1
0
1
2
2
1
, (1)B =

A= 3
2
2
3
0
2
4
3
3
3

2AB = 2

3 0

3C+D = 3 1

2
5B +

1
2006 D

2
0
5

1 2

6 0

1 2

=
2 4
4
0 4
2 4
3



1
3 6 9
1 6
2
3






3 = 3 0 12 + 2
4 + 2 1



1
6 15 3
1 2 4
1

is not defined.

14

2
3

6
1
2

3 = 1

7
4

12
1
17

15 .

Definition 1.3.5 Lat A be an m n matrix and B be an n p matrix. Then the product AB is

the m p matrix C whose entry cij in row i and column j is obtained as follows: Sum the products
formed by multiplying, in order, each entry in row i of the matrix A with the corresponding entry
in column j of the matrix B. More precisely
cij = ai1 b1j + ai2 b2j + + ain bnj =

n
X

aik bkj .

k=1

mp

must be
the same

np

mn

size of product

a11 a12 . . . a1n

a21 a22 . . . a2n

.
..
..
..
..
.
.
.

AB =
a
i1 ai2 . . . ain
.
..
..
..
.
.
.
.
.

am1 am2 . . . amn

n
n
X
X
a1k bk2
a
b

1k k1
k=1
k=1
n
n
X
X

a2k bk2
a
b

2k k1
k=1
k=1

..
..

.
.
=
n
n
X
X

aik bk1
aik bk2

k=1
k=1

..
..

.
.

n
n
X
X

amk bk1
amk bk2
k=1

b11

b12

. . . b1j

b21
..
.

b22
..
.

. . . b2j
.
..
. ..

bn1

bn2

. . . bnj

...
...
..

a1k bkj

k=1
n
X

a2k bkj

k=1

...
..

n
X

n
X
k=1

...

k=1

n
X

...
...

..
.

..

aik bkj

...

..
.

..

amk bkj

k=1

...

. . . b1p

. . . b2p

.. =
..
. .

. . . bnp
n
X

a1k bkp

a2k bkp

k=1

..

.
.
n

X
aik bkp

k=1

..

amk bkp
k=1
n
X

k=1

Definition 1.3.6 The transpose of an m n matrix A, denoted by AT , is the n m matrix whose

i th row is the i th column of A. Thus (AT )ij = (A)ji . Observe that for any matrix A we have
(AT )T = A.

Example 1.3.7 If A =

0 1

and B =
3 1 1 0

3
4
3
2 1

1 0 1
0 2

matrices AB, BA, (AB)T , (BA)T , AT B T , B T AT .

15

, find, when possible, the

Solution: then we have

AB =
|

0 1

3 1 1 0 =

3
3
2 1
} 4
|
{z
}

1 0 1
0 2
{z
23

34

1 2 + 0 3 + (1) 4

1 1 + 0 (1) + (1) 3

1 0 + 0 (1) + (1) 2

1 1 + 0 0 + (1) 1

02+23+34

0 1 + 2 (1) + 3 3

0 0 + 2 (1) + 3 2

01+20+31

=
|

2 18

2
7
.
, such that (AB)T =

2 4
4 3

}
0 3
|
{z
}

2 2 2 0
18

7
{z
24

42

Observe that the product

BA, (BA)T , AT B T

2
3

1 1
T
B =

0 1

1
0
{z
|

are not defined. However, observe that

1
0

and AT =
0 2 ,

1 3
1
{z
}
|
32
}

43

such that the product B T AT is also defined and

2
3 4

1 0

1
1
3
T T
0 2
B A =

0 1 2

1 3
1
0 1 |
{z
32
{z
}
|
43

2 18

2
7

= (AB)T .
=
2 4

0 3
}
{z
}
|
42

Definition 1.3.8 If A is a square matrix, then the trace of A, denoted by tr(A), is defined to be
the sum of the entries on the main diagonal, namely tr(A) := (A)11 + (A)22 + + (A)nn , where n
is the order of A. The trace of A is not defined if A is not a square matrix.
Example 1.3.9 Find, when possible, tr(AAT ) and tr(AB), where

2
1
0 1

1 0 1

and B =
A=
3 1 1 0

0 2
3
4
3
2 1

16


We first observe that AAT =

1 0 1
0 2

2
3

, such that tr(AAT ) =

0 2 =

3 13
1 3
|
{z
}
1 0

22

2 + 13 = 15. Since AB =

2 2 2 0
18

7
{z

4 3

is not a square matrix, its trace is not defined.

}

24

If Aij , 1 i m, 1 j n are matrices of suitable sizes, then we can form a new matrix

A=

A11

A12

A1n

A21

A22

A2n

Am1

Am2

Amn

and call partition of the matrix A the above kind of subdivision. For example

where

-1

-2

-1

-1

-1

-7

A11 =

1 0
1 0

A21 =

, A12 =

1 1
0 5

0 1

, A22 =

A13

A21

A22

A23

2 7

, A13 =

A12

A11

2 0

, A23 =

3 5

1 0

6 0

An application of matrix multiplication operation consists in transforming a system of linear equations into a matrix equation. Indeed, for a linear system we have successively:

a11 x1 + a12 x2 + . . . + a1n xn

= b1

a21 x1 + a22 x2 + . . . + a2n xn

..
.

= b2
..
.

am1 x1 + am2 x2 + . . . + amn xn

= bm

a11 x1 + a12 x2 + . . . + a1n xn

a21 x1 + a22 x2 + . . . + a2n xn
..
.
am1 x1 + am2 x2 + . . . + amn xn

17

b1
b2
..
.
bm

a11

a21

..
.

am1
|

a12

...

a1n

a22
..
.

...
..
.

a2n
..
.

x1
x2
..
.

am2 . . . amn
xn
{z
} | {z }
A

The matrix

b1
b2
..
.

AX = B.

bm
| {z }

a11

a21
A=
..
.

am1

a12

...

a1n

a22
..
.

...
..
.

a2n
..
.

am2

. . . amn

is called the coefficient matrix of the linear system. Observe that the augmented matrix of the
h
i
.
system is A .. B .

1.4

Inverses. Rules of matrix arithmetic

Theorem 1.4.1 Assuming that the sizes of matrices are such that the indicated operations can be
performed, the following rules of matrix arithmetic are valid:
1. A + B = B + A;
2. (A + B) + C = A + (B + C);
3. (AB)C = A(BC);
4. A(B + C) = AB + AC;
5. (A + B)C = AC + BC;
6. A(B C) = AB AC;
7. (A B)C = AC BC;
8. a(B + C) = aB + aC;
9. a(B C) = aB aC;
10. (a + b)C = aC + bC;
11. (a b)C = aC bC;
12. a(bC) = (ab)C;

18

13. a(BC) = (aB)C = B(aC).

Remark 1.4.2 The matrix multiplication is not commutative. Indeed, for example if

1 0
1 2
, B =
,
A=
2 3
3 0
then

AB =

1 0
2 3

1 2
3 0

1 2
11

6=

3 6

3 0

1 2

Define the m n zero matrix to be the matrix Omn

1 0

3 0

:=

= BA.

2 3

0 0 0
0
..
.

0
.. . .
.
.

0
..
.

0 0 0
denoted by O when the size is understood from context.

and it will be equally

mn

Theorem 1.4.3 Assuming that the sizes of the matrices are such that the indicated operations can
be performed, the following rules of matrix arithmetic are valid
1. A + O = O + A = A;
2. A A = O;
3. O A = A;
4. AO = O.
The identity matrix of order n is defined to be the square the n n matrix

1 0 0

0 1 0

,
In :=
.. .. . . ..
. .
. .

0 0 1
nn

and it will be equally denoted by I when the order is understood from context.
Remark 1.4.4 If A is an m n matrix, then AIn = Im A = A.
Theorem 1.4.5 If R is the reduced row-echelon form of an n n matrix A, then either R has a
row of zeros, or R is the identity matrix.

19

Definition 1.4.6 A square matrix A is said to be invertible if there is another square matrix B,
of the same size, such that AB = BA = I.
Remark 1.4.7 If B, C are both inverses of a A, then B = C. Indeed, we have successively:
B = BI = B(AC) = (BA)C = IC = C. Therefore, the inverse of an invertible matrix A is unique
and denoted by A1 , being characterized by the equalities: AA1 = A1 A = I.

Example 1.4.8 If a, b, c, d R are such that ad bc 6= 0, then the matrix A =

invertible and A1 =

1
adbc

a b
c

{z

d b
c

d
adbc

b
adbc

c
adbc

a
adbc

d
adbc

b
adbc

c
adbc

a
adbc

a b
c

is

. Indeed we have

adbc
adbc

ab+ba
adbc

cddc
adbc

cb+da
adbc

1 0

0 1

and similarly

d
adbc

b
adbc

c
adbc

a
adbc

a b
c

{z

dabc
adbc

dbbd
adbc

ca+ac
adbc

cb+ad
adbc

1 0

0 1

If A is a square matrix, then its positive powers are defined to be A0 = I, An = AA

A} for n > 0.
| {z
1 1
1
Moreover, if A is invertible, then An = A
| A {z A }, also for n > 0.

n times

n times

Theorem 1.4.9 If A is an invertible matrix, then

1. A1 is also invertible and (A1 )1 = A;
2. An is invertible and (An )1 = (A1 )n for n {0, 1, 2, . . .};
3. AT is invertible and (AT )1 = (A1 )T ;
4. For any non-zero scalar k, the matrix kA is invertible and (kA)1 = k1 A1 ;
5. Am An = Am+n for any integers m, n;
6. (Am )n = Amn for any integers m, n.

Theorem 1.4.10 If the sizes of the involved matrices are such that the stated operations can be
performed, then

20

1. (AT )T = A;

2. (A + B)T = AT + B T and (A B)T = AT B T ;

3. (kA)T = kAT , where k is any scalar;

4. (AB)T = B T AT .

Example 1.4.11

1. Find the matrix A knowing that (I2 + 2A)1 =

Solution: Since (I2 + 2A)1 =

1 2

4 5

1 2
4 5

1
1 2

, it follows that [(I2 + 2A)1 ]1 =

4 5

Therefore we have successively

5 2
5 2
1

I2 + 2A = 1

(1) 5 2 4 4 1
13 4 1

I2 + 2A =

2A =

5
13

1 5 2 1 0

2A =
13 4 1
0 1

2A =

A=

cos

sin

sin

cos

18
13

2
13

4
13

12
13

12
1
2

18
13

4
13

1
2

A= 1
2

2
13

12

2. If A =

12
13

4
13

12
13

A=

sin 2

sin 2

cos 2

9
13

1
13

2
13

6
13

cos 2

1
13

4
13

2
13

, show that A2 =

18
13

2
13

and A3 =

cos 3

sin 3

sin 3

cos 3

Solution: Indeed,
A2 =

cos

sin

sin

cos

cos

sin

sin

cos

21

cos2 sin2

2 sin cos

2 sin cos

cos2 sin2

cos 2

sin 2

sin 2

cos 2


A3 = AA2 =

cos sin
sin

cos

cos 2 sin 2
sin 2

cos 2

cos cos 2 sin sin 2 (cos sin 2 + sin cos 2)

sin cos 2 + cos sin 2

cos( + 2) sin( + 2)
sin( + 2)

cos 3 sin 3
sin 3

1.5

cos( + 2)

sin sin 2 + cos cos 2

cos 3

Elementary matrices and a method for finding A1

Definition 1.5.1 An n n matrix is called an elementary matrix if it can be obtain from the n n
identity matrix In by performing a single elementary row operation.
Example 1.5.2

I3 = 0 1 0

0 0 1

If A =

1 0 0

3r + r r

1
3
3
0 1 0

3 0 1

1 0 0

3 , then EA =

= E elementary matrix.

r r +r
3 3 3 1

3 = A.

Theorem 1.5.3 If the elementary matrix E results from performing a certain row operation on
Im and A is an m n matrix, then the product EA is the matrix that results when the same row
operation is performed on A.
If an elementary operation is performed on the identity matrix I to obtain an elementary matrix
E, then there is another operation, called the corresponding inverse operation, when apply to E to
obtain I back again.
Direct Operation

Corresponding Inverse operation

Multiply row i by c 6= 0 (cri )

Multiply row i by

1
c

( 1c ri )

Interchange rows i and j (i j)

Interchange rows i and j (i j)

Add c times row i to row j (cri + rj rj )

Add c times row i to row j (cri + rj rj )

22

Theorem 1.5.4 Every elementary matrix is invertible, and the inverse is also an elementary matrix.
dir. op.

inv. op.

inv. op.

Proof. If I E I. Daca I E0 , then E0 E = I and EE0 = I, namely E is

invertible.
Theorem 1.5.5 If A is an n n matrix, then the following statements are equivalent:
1. A is invertible;
2. The homogeneous linear system AX = O has only the trivial solution;
3. The reduced row-echelon form of A is In ;
4. A is expressible as a product of elementary matrices.

The last statement, (4), of theorem 1.6.7 is of particular importance since it provide us a method
of finding the reduced row-echelon form of a matrix and a method of finding the inverse of an
invertible matrix.
Let A be a matrix and R be its reduced row-echelon form, namely
op

op

op

op

A 1 A1 2 A2 3 n R,
op

op

op

then R = En . . . E1 A, where I 1 E1 , I 2 E2 , , I n En . Therefore

A = (En . . . E1 )1 R = E11 . . . En1 R.
If A is particularly invertible, then R = I, such that A = E11 . . . En1 and implicitly
A1 = En . . . E1 .
Thus, in order to find the inverse of an invertible matrix A, we first observe that we have the general
.
.
property A[B ..C] = [AB ..AC] and that we have successively:
op
op
. op
.
.
.
.
.
op
[A..I] 1 E1 [A..I] = [E1 A..E1 I] 2 E2 [E1 A..E1 I] = [E2 E1 A..E2 E1 I] 3 n [En E1 A .. En E1 I ].
| {z } | {z }
=I

Examples 1.5.6

1. Find the inverse of the matrix A =

2 1
3
2

product of elementary matrices;

23

=A1

1 7 and write A1 as a

0 5

2. Express the matrix

A=

1 7

3 3

2 5 1 8

in the form A = EF GR, where E, F, G are elementary matrices and R in row-echelon form.
(1)

2 1
3

4 1 0 0
1 0 3

r +r r

1
1
3 1 7
7 0 1 0

5
0 0 1
2 0 5

1
1 0
0 3
1

2r + r r

3
2
2
0
1
2 3 2 0

3r3 + r1 r1

0
1 2 2 1
0

1 1 0

3r + r r

1
2
2

0 1 0

2r1 + r3 r3

0 0 1

0 0 5 5
3

1 0
1
2 2 .

0 1 2 2
1
|
{z
}
=A1

Therefore E5 E4 E3 E2 E1 A = I3 , where

1 0 0
1 1 0

r +r r

1
1
I3 = 0 1 0 2
0 1 0 =: E1

0 0 1
0 0 1

1 0 0
1 0 0

3r + r r

1
2
2
3 1 0 =: E2
I3 = 0 1 0

0 0 1
0 0 1

1 0 0
1 0 0

2r + r r

1
3
3
0 1 0 =: E3
I3 = 0 1 0

2 0 1
0 0 1

1 0
0
1 0 0

2r + r r

3
2
2
0 1 2 =: E4
I3 = 0 1 0

0 0
1
0 0 1

1 0 0
1 0 3

3r + r r

1
1
I3 = 0 1 0
0 1 0 =: E5 .

0 0 1
0 0 1
(2)

A=

1 7

3 3

3 3

r r

2
0
8
1 7
8

2 5 1 8
2 5 1 8

24

2r + r r
1
3
3


2r1 + r3 r3

1 3 3 8

0 1 7 8

0 1 7 8

1 3 3 8
r + r r

2
3
3
0 1 7 8

0 0 0 0

= Rmatrix in row-echelon form.

Consequently R = E3 E2 E1 A, or equivalently A = E11 E21 E31 R, where

1 0 0
0 1 0
r r

2
I3 = 0 1 0 1
1 0 0 =: E1 = E11

0 0 1
0 0 1

1 0 0

I3 = 0 1 0

0 0 1

2r1 + r3 r3 %

1 0 0
0 1 0
2 0 1

=:E

2r1 + r3

r3 &

1 0 0
0 1 0
2 0 1

1 0 0

I3 = 0 1 0

0 0 1

r2 + r3

r3 %

0 0

1 0

0 1 1

r2 + r3

=:E
3

r3 &

=:E 1
2

1 0 0
0 1 0
0 1 1

=:E 1
3

Now it is enough to take E = E11 , F = E21 , G = E31 .

1.6

Further results on systems of equations and invertibility

Theorem 1.6.1 Every system of linear equations has either no solution, exactly one solution or
infinitely many solutions.
Theorem 1.6.2 If A is an invertible n n matrix, then for each n 1 matrix b, the system of
linear equations AX = B has exactly one solution, namely X = A1 B.
Example 1.6.3 Find the solution set of the following linear system:
2x y + 4z =

3x + y 7z = 1
2x

5z =

25

 Solution: We first observe that the coefficient matrix of the given linear
system is A =
2 1
4
5 5
3

3
1 7 , which is invertible and its inverse is A1 = 1
2 2 . The given linear

2
0 5
2 2
1

x
1

system can be written in the form AX = B, where X = y and B = 1 . Consequently the

z
0

0
1
5 5
3

unique solution of the given linear system is x = A1 B = 1

2 2 1 = 1 .

0
0
2 2
1
Frequently, one have to solve a sequence of linear systems
AX = B1 , AX = B2 , . . . , AX = Bk
each of which has the same square matrix A. If A is invertible then the solutions are
X1 = A1 B1 , X2 = A1 B2 , . . . , Xk = A1 Bk .
Otherwise the method of solving those systems, which works both for A invertible or A noninvertible, consists in forming the matrix
.
.
.
.
[A .. B1 .. B2 .. .. Bk ]
and by reducing it to the reduced row-echelon form.
Example 1.6.4 Solve the following linear systems:
2x y + 4z =

2x y + 4z =

3x + y 7z = 1 ,

3x + y 7z = 2

2x

2x

5z =

5z =

Theorem 1.6.5 Let A be a square matrix.

1. If B is a square matrix satisfying BA = I, then B = A1 ;
2. If B is a square matrix satisfying AB = I, then B = A1 .
Theorem 1.6.6 Theorem 1.6.7 If A is an nn matrix, then the following statements are equivalent:
1. A is invertible (*);

26

2. The homogeneous linear system AX = O has only the trivial solution;

3. The reduced row-echelon form of A is In ;
4. A is expressible as a product of elementary matrices.
5. AX = B is consistent for every n 1 matrix B;(*)
6. AX = B has exactly one solution for every n 1 matrix B(*).
Theorem 1.6.8 Let A and B be square matrices of the same size. If AB is invertible, then A and
B are both invertible.
Fundamental Problem 1.6.9 Let A be a fixed m n matrix. Find all m 1 matrices B such
that the system of equations AX = B is consistent.
Example 1.6.10 What conditions must a, b1 , b2 , b3 satisfy in order for the system of the linear
system
x +

y 2z = b1

x 2y +
2x +

y +

z = b2
z = b3

to be consistent ?
Solution: For the augmented matrix of the given linear system we have

.
.
1
1 2 .. b1
1
1
a ..
b1

r + r r
.
.

1
2
2
0 3
1 2
1 .. b2
3 .. b2 b1

2r1 + r3 r3
.
.
2
1
1 .. b3
0
3 8 .. b3 + 2b1

.
1
1 a ..
b1

r1 + r2 r2
.

0 3 3 ..
.
b2 b1

..
0
0 0 . b3 + b2 + b1

successively:

r +r r
1
2
2

Thus, the given linear system is consistent if and only if b3 + b2 + b1 = 0.

1.7

Diagonal, Triangular and Symmetric Matrices

Definition 1.7.1 A square matrix in which all the entries off the diagonal are zero is called a
diagonal matrix. The general form a of a diagonal

d
1

0 d2
D=
..
..
.
.

27

matrix is
0

..
.

0
..
.

dn

Remarks 1.7.2

1. The diagonal matrix

D=

d1

0
..
.

d2
..
.

..
.

dn

is invertible iff all of its diagonal entries are non-zero.

1
0
d1

1
0

d2
1

D = .
.
..
..
..
.

0 0
2. Powers of diagonal matrices are easy to

d 0 0
1

0 d2 0
D=
..
..
. . ..
.
. .
.

0
3. If A = [aij ]

..
.

In this case

.. .
.

1
dn

be computed. More

dk

0
then Dk =

..

dn

is an m n matrix, then

0 0
a
a12
11

2 0 a21 a22

.
..
..
. . ..
. . ..
.
.

am1

am2

0
..
.

...

a1n

...
..
.

a2n
..
.

. . . amn

precisely, if

dk2
..
..
.
.
0

0
..
.

dkn

1 a11

2 a21
=

..

.

m am1

1 a12

...

1 a1n

2 a22
..
.

...
..
.

2 a2n
..
.

m am2

. . . m amn

and

a11

a21

..
.

am1

a12

...

a1n

a22
..
.

...
..
.

a2n
..
.

am2

. . . amn

d1
0
..
.

d2
..
.

a11 d1

0 a21 d1
=

..
. . ..
. .
.

am1 d1
dn

Examples 1.7.3 Find a diagonal matrix A satisfying

9 0 0

2
A = 0 4 0

0 0 1

28

a12 d2

...

a1n dn

a22 d2
..
.

...
..
.

a2n dn
..
.

am2 d2

. . . amn dn

Solution: We have successively:

9 0 0
9 0 0

2 1
= 0 4 0 [A ] = 0 4 0

0 0 1
0 0 1

1
9

A2 = 0

0 0

1
3

2
0 A =

0 1
1
4

Thus, such a matrix is

0
2
1
2

0
0

0 .

2
1

1
3

0 0

A= 0

0 .

0 1
1
2

Definition 1.7.4 A square matrix in which all the entries above the main diagonal are zero is
called lower triangular and a square matrix in which all the entries below the main diagonal are
zero is called upper triangular. A matrix which is either lower triangular or upper triangular is
called triangular. The general form a of a lower triangular matrix is

a11

a21

..
.

an1

a22
..
.

..
.

0
..
.

an2

ann

and the general form a of a upper triangular matrix is

Remark 1.7.5

a11

a12

a1n

0
..
.

a22
..
.

a2n
..
..
.
.

ann

1. A square matrix A = [aij ] is upper triangular iff aij = 0 for i > j;

2. A square matrix A = [aij ] is lower triangular iff aij = 0 for i < j.

Theorem 1.7.6

1. The transpose of a lower triangular matrix is upper triangular, and the

transpose of an upper triangular matrix is lower triangular;

2. The product of lower triangular matrices is a lower triangular matrix;

29

3. The product of upper triangular matrices is a upper triangular matrix;

4. A triangular matrix is invertible if and only if its diagonal entries are all nonzero;
5. The inverse of a lower triangular matrix is lower triangular and the inverse of an upper
triangular matrix is upper triangular.
Example 1.7.7 If

A= 0

2
0

1 1 2

3 and B = 0 1 2

0 0 5
1

show that AB is also upper triangular and find (AB)1 and B 1 A1 .

Solution:

1 1 2

9 1 0

0 2 19 0 1

0
0 5 0 0

1 0 92

0 1 0

0 0 1
2

3 1

0
2 3 0 1

0
0 1
0
0

1R , 1R , 1R
2 1 2 2 5 3

1
1
0
2
9R + R
2 3
1

0 12 19
10
0
0 15

2 = 0 2

5
0
0

9 1
1 2
2 2

0 1 19
2 0

0 0
1 0

1 0
R1

0 1

0 0

Thus

1
2

9
1 10

(AB)1 = B 1 A1 = 0 12

0
0

19
10

19 .

0
21

2R + R R

2
1
1

0 19
2 R3 + R2 R2

0 15

9
0 21
1 10

19
0 0 21
10

1
0
1 0
5

1
5

Definition 1.7.8 A square matrix A is called symmetric if AT = A, or equivalently (A)ij = (A)ji

Theorem 1.7.9 If A, B are symmetric matrices of the same size and k is any scalar, then:
1. AT is symmetric;
2. A + B and A B are symmetric;
3. kA is symmetric.
Theorem 1.7.10

1. The product of two symmetric matrices is symmetric if and only if the

matrices commute;

30

2. If A is an invertible symmetric matrix, then A1 is symmetric.

Theorem 1.7.11

1. If A is an arbitrary matrix, then AAT and AT A are symmetric matrices;

2. If A is an invertible matrix, then AAT and AT A are invertible matrices.

Example 1.7.12 Find the values of a, b, c R for which the

2 a + b + c 4b 3c

A= 5
1
2a + c

2
4
6

matrix A is symmetric, where

a+b+c=5
Solution: The required values are solutions of the linear system

4b 3c = 2

that

2a + c = 4
can be solved by performing row-reduction operations in the corresponding augu

.
.
1 1 1 .. 5
1 1
1 .. 5

1 r2 1 r2
.
.

2r1 +r3 r3
4
4
mented matrix, namely: 0 4 3 .. 2

0 4 3 .. 2

.
.
2 0 1 .. 4
0 2 1 .. 6

.
.
.
1 1
1 .. 5
1 0 74 .. 92
1 1 1 .. 5

2 r3

.
.
.

2r2 +r3 r3
5

0 1 34 .. 12
0 1 34 .. 12 0 1 34 .. 12 . The corre

..
..
5 ..
0 2 1 . 6
0 0 2 . 5
0 0 1 . 2
sponding linear system of the last matrix in the above row-reduction process, which is equivalent
with the initial one, has the unique solution a = 1, b = c = 2.

31

32

Chapter 2

Determinants
2.1

The determinant function

Definition 2.1.1 A permutation of the set {1, 2, . . . , n} is a bijective function : {1, 2, . . . , n}

{1, 2, . . . , n}, namely an arrangement of the integers 1, 2, . . . , n without omissions or repetitions. It
is also written

(1) (2) (n)

An inversion of the permutation is a pair (i, j) such that i < j and (i) > (j). Denote by m()
the number of inversions of and by () the signature (1)() of . A permutation is called
even/odd if m() is even/odd, or equivalently () = +1/() = 1.
Definition 2.1.2 The determinant of a square n n matrix A = [aij ] is defined to be
det(A) =

()a1(1) a2(2) . . . an(n) .

It is denoted either by det(A) or by

a11

a21

..
.

an1

Remarks 2.1.3

1. If A =

a12
a22
..
.
an2

. . . a1n

. . . a2n

.. .
..
.
.

. . . ann

a11

a12

a21

a22

is a 2 2 square matrix, then det(A) = a11 a22 a12 a21 .

Indeed, the only 2 2 permutations are e =

1 2
1 2

33

and =

1 2
2 1

and (e) = 1,

() = 1 since m(e) = 0 and m() = 1. Thus

a11 a12
a11 a12
= (e)a

a
+ ()a1(1) a2(2) = a11 a22 a12 a21 .
det
=
1e(1) 2e(2)

a21 a22
a21 a22

1 2

= 1 3 (1)3 = 3 + 3 = 6.
For example

1 3

2. A 2 2 matrix A =

a11

a12

a21

a22

is invertible if and only if det(A) 6= 0. In this case

A1

1 a22
=
det(A) a
21

a12

a11

a11

a12

a13

3. If A = a21 a22 a23 is a 2 2 square matrix, then

a31 a32 a33

a11 a12 a13

det(A) = a21 a22 a23 = a11 a22 a33 +a12 a23 a31 +a13 a21 a32 a13 a22 a31 a12 a21 a33 a11 a23 a32 =

a31 a32 a33

a11

a12

a13

a11

a12

a21

a22

a23

a21

a22

a31

a32

a33

a31

Indeed, the only 3 3 square matrices are

1 2 3
1
, 1 :=
e=
1 2 3
2

1 2 3
1
, 4 =
3 =
1 3 2
2

2 3
1 3

3 1

, 2 =

2 3

a32

1 2 3

, 5 =

3 2 1

1 2 3

3 1 2

and (e) = (4 ) = (5 ) = 1 while (1 ) = (2 ) = (4 ) = 1 since m(e) = 0, m(4 ) =

m(5 ) = 2 and m(1 ) = 1, m(2 ) = 3, m(3 ) = 1.

1 1 2 1 1 2
1 1

For example 0
0
2 = 6 + 2 + 0 (8) 2 0 = 8 + 8 = 16.
2 1
2 1 = 0

2
2
2 3 2
2 3 2

34

2.2

Evaluating determinants by row reduction

Theorem 2.2.1 Let A be a square matrix.

1. If A has a row of zeros, then det(A) = 0.
2. det(A) = det(AT ).
Theorem 2.2.2 If A is an n n triangular matrix, then det(A) is the product of the entries on
the main diagonal, namely det(A) = a11 a22 . . . ann
Theorem 2.2.3 Let A be an n n square matrix.
1. If B is the matrix that results when a single row or a single column of A is multiplied by a
scalar k, then det(B) = k det(A); Consequently det(kA) = k n det(A).
2. If B is the matrix that results when two rows or two columns of A are interchanged, then
det(B) = det(A);
3. If B is the matrix that results when a multiple of one row of A is added to another row or
when a multiple of one column of A is added to another column, then det(B) = det(A).
Corollary 2.2.4 If A is a square matrix with two proportional rows or two proportional columns,
then (A) = 0.

1 2 3

1 0 1
0
Example 2.2.5 Find the determinant of A =

0 1 0
1

0 0 1 1
Solution:

det(A)=

4
= (2)

1 2 3

1 0 1

0 1 0

0 0 1 1
r1 + r2 r2

2
= (2) 1

1
0
1

1 =

= 2(2 1 + 1) = 4.

35

2.3

Properties of determinant function

Theorem 2.3.1 Let A, B and C be n n matrices that differ only in a single row, say the rth , and assume
that the rth row of C can be obtained by adding corresponding entries in the rth row of A and B. Then
det(C) = det(A) + det(B).
In other words

a11

a21

..

a11 + a011

..

am1

a12

a1n

a22
..
.

..
.

a2n
..
.

a12 + a012
..
.

..
.

a1n + a01n
..
.

am2

amn

a11

a12

a21
..
.

a22
..
.

..
.

a11
..
.

a12
..
.

..
.

am1

am2

a1n

a2n
..
.
+

a1n

..
.


a
mn

a11

a12

a21
..
.

a22
..
.

..
.

a011
..
.

a012
..
.

..
.

am1

am2

a1n

a2n
..
.

a01n

..
.

a
mn

Theorem 2.3.2 If A and B are square matrices of the same size, then det(AB) = det(A) det(B).
Theorem 2.3.3 If A is an invertible matrix, then det(A1 ) =

1
det(A) .

Theorem 2.3.4 If A is an n n matrix, then the following statements are equivalent:

1. A is invertible;(*)
2. The homogeneous linear system AX = O has only the trivial solution;
3. The reduced row-echelon form of A is In ;
4. A is expressible as a product of elementary matrices.
5. AX = B is consistent for every n 1 matrix B;
6. AX = B has exactly one solution for every n 1 matrix B.
7. det(A) 6= 0(*)

Example 2.3.5

1. Consider the matrix A =

y+z

z+w

. Show that det(A) = 0

w+x x+y

for all x, y, z, w R.
Solution: For the determinant of A we have successively:

1
1
1
1

x
y
z
w r2 +r4 r4
x

x
y
z
w

y+z z+w w+x x+y

x+y+z

36

1
y
x
y+z+w

z
w

y
z

z+w+x w+x+y
1

x+y+z+w

r3 +r4 r4

z
w
=

y
z

y+z+w+x z+w+x+y
1

y
x
x+y+z+w

= (x + y + z + w)

x y

w
=0

for all x, y, z, w R.

2. Let

A= d

f .

h i
e

Assuming that det(A) = 7, find det(3A), det(A1 ), det(2A1 ) and det b

g
h
i

e .

Solution: det(3A) = 33 det(A) = 27(7).

det(A1 ) =

1
det(A)

1
7

= 17 .

det(2A1 ) = 23 det(A1 ) = 8 17 = 87 .

det b

g
h
i

e = det b

c
f

a1 + b1 t

3. Show that a1 t + b1

c1

g
h
i

a b c

= g h i

d e f

= d

a1
a2 + b2 t a3 + b3 t

2
a2 t + b2 a3 t + b3 = (1 t ) b1

c1
c2
c3

37

a2
b2
c2

e f = (7) = 7.

h i
b

a3

b 3 .

c3

Solution: Indeed, we have successively:

a1 + b1 t

a1 t + b1

c1

a2 + b2 t
a2 t + b2
c2

a3 + b3 t

a3 t + b 3

c3

a1

a1 t + b1

c1

a1

a1 t

c1
|

a2

a3

a2 t + b2

a3 t + b3

c2

c3

a3 a1


a3 t + b1


c3 c1
}

a2
a2 t
c2
{z

a2
b2
c2

b1 t


+ a1 t + b1



c1


a3 b1 t


b3 + a1 t


c3 c1

b2 t

b3 t

a2 t + b2

a3 t + b3

c2

c3

b2 t
a2 t
c2

b3 t b1 t


a3 t + b1


c3 c1
|

=0

2.4

a1

b1

c1

a1

b1

c1

b2 t
b2
c2
{z
=0

a2

a3

b2

b3

c2

c3

b1

+ t2 a1

c1

a1
a3

2
b3 t b1

c1
c3

a2
b2
c2

a1

2
(1 t ) b1

c1

a2
b2
c2

b2

b3

a2

a3

c2

c3

a2
b2
c2

a3

b3

c3

a3

b3

c3

Cofactor expansion. Cramers rule

Definition 2.4.1 If A is a square matrix, then the minor of entry aij is denoted by Mij and is defined to
be the determinant of the submatrix that remains after the ith row and the j th column are deleted from A.
The number (1)i+j Mij is denoted by Cij and is called the cofactor of entry aij .

Example 2.4.2

a11

a21

a31

a12
a22
a32

a13

a23

a33

a11 a22 a33 + a12 a23 a31 + a13 a21 a32 a13 a22 a31 a12 a21 a33 a11 a23 a32

a11 (a22 a33 a23 a32 ) a12 (a12 a21 a33 a23 a31 ) + a13 (a21 a32 a22 a31 )

a22
= (1)1+1 a11
a32
=

a
a23
+ (1)1+2 a12 21

a31
a33

a11 C11 + a12 C12 + a13 C13

38

a
a23
+ (1)1+3 a13 21

a31
a33

a22

a32

b3 t

b3

c3
}

Similarly

a11

a21

a31

a12
a22
a32

a13

a23

a33

a11 a22 a33 + a12 a23 a31 + a13 a21 a32 a13 a22 a31 a12 a21 a33 a11 a23 a32

a12 (a21 a33 a23 a31 ) + a22 (a11 a33 a13 a31 ) a32 (a11 a23 a13 a21 )

a
1+2 21
= (1)

a31
=

a
a23
+ (1)2+2 11

a31
a33

a
a13
+ (1)3+2 11

a21
a33

a13

a23

a12 C12 + a22 C22 + a32 C32

Theorem 2.4.3 If A = [aij ] is a square n n matrix, then

det(A) =

ai1 Ci1 + ai2 Ci2 + + ain Cin

|
{z
}

The cofactor expansion of det(A) along the ith row

Example 2.4.4 Show

1 1

a1 a2

2
a1 a22

3
a1 a32

a1j C1j + a2j C2j + + anj Cnj

{z
}
|

The cofactor expansion of det(A) along the j th column

that:
1
a3
a23
a33

a4
= (a1 a2 )(a1 a3 )(a1 a4 )(a2 a3 )(a2 a4 )(a3 a4 )

a24

a34

If A = [aij ] is a square n n matrix and Cij is the cofactor of the entry aij , then the matrix

C11

C21

..
.

Cn1

C12

C1n

C22
..
.

..
.

C2n
..
.

Cn2

Cnn

is called the matrix of cofactors from A. The transpose of this matrix is called the adjoint of A and is
denoted by adj(A).

If A is an invertible matrix, then A1 =

1
det(A) adj(A).

The inverse of an invertible

lower triangular matrix is lower triangular and the inverse of an invertible upper triangular matrix is upper
triangular.

Find necessary and sufficient conditions

A= a

a2

on a, b, c R

1 1

a a

2
2
a a

such that the matrix A is invertible. In those cases find A1 .

39

According to theorem 2.3.4, A is invertible if and only if det(A) 6= 0. But since

det(A) = V (a, b, c) = a

2
a

1
b
b2

c = (c a)(c b)(b a),

c2

it follows that A is invertible if and only if a 6= b, b 6= c and c 6= a. In this case

T
C11

A1 =

1
1

adj(A) =
C
det(A)
det(A) 21
C31

2
b

c2

1
1
1

(c a)(c b)(b a) b2 c2

1 1

b c

C12

C13

C32

C23 =

C33

a

a2

c2

2
a

b2

2
a

c2

1

a2

b2

C22

1 1

a c

1 1

a b

b2

c2

bc(c b)

cb

=
ac(a c) c2 a2 a c
(c a)(c b)(b a)
ab(b a) a2 b2 b a

bc(c b) ac(c a) ab(b a)

=
(c2 b2 )
c2 a2
(b2 a2 )
(c a)(c b)(b a)
cb
(c a)
ba

bc(cb)
(ca)(cb)(ba)
ac(ac)
(ca)(cb)(ba)
ab(ba)
(ca)(cb)(ba)

bc
(ca)(ba)

b+c
(ac)(ba)

1
(ca)(ba)

ac
(cb)(ab)

c+a
(cb)(ba)

1
(cb)(ab)

ab
(ca)(cb)

a+b
(ca)(bc)

1
(ca)(cb)

(bc)(b+c)
(ca)(cb)(ba)
(ca)(c+a)
(ca)(cb)(ba)
(ab)(a+b)
(ca)(cb)(ba)

cb
(ca)(cb)(ba)
ac
(ca)(cb)(ba)
ba
(ca)(cb)(ba)

(Cramers Rule) If AX = B is a system of n linear equations in n unknowns such that det(A) 6= 0,

then the system has a unique solution. This solution s
x1 =

det(A1 )
det(A2 )
det(An )
, x2 =
, . . . , xn =
,
det(A)
det(A)
det(A)

where Aj is the matrix obtained from A by replacing its j th row with the entries of the column
matrix B.

Provide necessary and sufficient conditions on a1 , a2 , a3 , a4 R such that the linear

40

system
x +

y +

z +

=1

ax +

by +

cz +

dw

a2 x + b2 y + c2 z + d2 w = 2

(2.1)

a3 x + b3 y + c3 z + d3 w = 3
has exactly one solution for each R. In this case solve the system by using the Cramers rule.
The given linear system has a unique solution for each R if and only if its coefficient matrix

1 1 1 1

a b c d

A=

2
2
2
2
a b c d

3
3
3
3
a b c d
is invertible, or, equivalently det(A) 6= 0. But since
det(A) = V (a, b, c, d) = (d a)(d b)(d c)(c a)(c b)(b a),
it follows that the given linear system has unique solution for each R if and only if the scalars
a, b, c, d are pairwise disjoint.
By Crammers rule, it follows that the unique solution of the system is
x=

V (, b, c, d)
(d )(c )(b )
=
;
V (a, b, c, d)
(d a)(c a)(b a)

y=

( a)(c )(d )
V (a, , c, d)
=
;
V (a, b, c, d)
(b a)(c b)(d b)

z=

V (a, b, , d)
( a)( b)(d )
=
;
V (a, b, c, d)
(c a)(c b)(d c)

w=

V (a, b, c, )
( a)( b)( c)
=
.
V (a, b, c, d)
(d a)(d b)(d c)

41

42

Chapter 3

Euclidean vector spaces

3.1

Euclidean n-space

Definition 3.1.1 If n is a positive integer, then an ordered n-tuple is a sequence of n numbers

(a1 , a2 , . . . , an ). The set of Rn of all n-tuples is called n-space.
A 2-tuple is also called ordered pair and a 3-tuple is called ordered triple and both of them have
geometric interpretation. Thus an n-tuple can be viewed as either an generalized point or a
generalized vector.
Definition 3.1.2 Two vectors u = (u1 , u2 , . . . , un ) and v = (v1 , v2 , . . . , vn ) in Rn are called equal
if u1 = v1 , u2 = v2 , . . . , un = vn . Their sum u + v is defined by u + v = (u1 + v1 , u2 + v2 , . . . , un + vn )
and, for a sscalar k, the scalar multiple is defined by ku = (ku1 , ku2 , . . . , kun ).
The zerovector in Rn is denoted by 0 and is defined to be the vector 0 = (0, 0, . . . , 0). If u =
(u1 , u2 , . . . , un ) Rn is any vector, the negative or additive inverse of u is denoted by u and
is defined by u = (u1 , u2 , . . . , un ). The difference u v of the vectors (u1 , u2 , . . . , un ), v =
(v1 , v2 , . . . , vn ) Rn is defined by u v := u + (v). In termes of components we have u v =
(u1 v1 , u2 v2 , . . . , un vn ).
Theorem 3.1.3 If u = (u1 , u2 , . . . , un ), v = (v1 , v2 , . . . , vn ), w = (w1 , w2 , . . . , wn ) Rn and k, l are
scalars, then
1. u + v = v + u.
2. u + (v + w) = (u + v) + w.
3. u + 0 = 0 + u = u.
4. u + (u) = 0, namely u u = 0.

43

5. k(lu) = (kl)u.
6. k(u + v) = ku + kv.
7. (k + l)u = ku + lu.
8. 1u = u.
Definition 3.1.4 If u = (u1 , u2 , . . . , un ), v = (v1 , v2 , . . . , vn ) Rn are any vectors, then the Euclidean Product u v is defined by u v := u1 v1 + u2 v2 + + un vn .
Theorem 3.1.5 If u = (u1 , u2 , . . . , un ), v = (v1 , v2 , . . . , vn ), w = (w1 , w2 , . . . , wn ) Rn and k, l are
scalars, then
1. u v = v u.
2. k(u v) = (ku) v = u (kv).
3. (u + v) w = u w + v w.
4. u u 0 and u u = 0 if and only if u 0.

Definition 3.1.6 The Euclidean norm or the Euclidean length ||u|| of a vector u =
p

(u1 , u2 , . . . , un ) Rn is defined by ||u|| := u u = u21 + u22 + + u2n and the distance d(u, v)
between the points u = (u1 , u2 , . . . , un ), v = (v1 , v2 , . . . , vn ) Rn is defined by
d(u, v) = ||u v|| =

p
(u1 v1 )2 + (u2 v2 )2 + + (un vn )2 .

Theorem 3.1.7 (Cauchy-Schwarz Inequality in Rn ) If u = (u1 , u2 , . . . , un ), v = (v1 , v2 , . . . , vn )

Rn , then |u v| ||u|| ||v||. In terms of components, the inequality becomes
q
|u1 v1 + u2 v2 + + un vn |

q
u21 + u22 + + u2n

Theorem 3.1.8 If u, v Rn and k is any scalar, then

1. ||u|| 0.
2. ||u|| = 0 if and only if u = 0.
3. ||ku|| = |k| ||u||.

44

v12 + v22 + + vn2 .

4. ||u + v|| ||u|| + ||v|| (Triangle inequality).

Theorem 3.1.9 If u, v, w Rn and k is any scalar, then
1. d(u, v) 0.
2. d(u, v) = 0 if and only if u = v.
3. d(u, v) = d(v, u).
4. d(u, v) d(u, w) + d(w, v) (Triangle inequality).
Theorem 3.1.10 If u, v Rn , then u v = 14 ||u + v||2 41 ||u v||2 .
Proof. Indeed, by adding the identities
||u + v||2 = (u + v) (u + v) = ||u||2 + 2u v + ||v||2
||u v||2 = (u v) (u v) = ||u||2 2u v + ||v||2 ,
we immediately get the required identity.
Definition 3.1.11 Two vectors u, v Rn are said to be orthogonal if u v = 0.
Theorem 3.1.12 (Theorem of Pythagoras for Rn ) If u, v Rn are orthogonal vectors, then the
equality ||u + v||2 = ||u||2 + ||v||2 holds.
Proof. Indeed, we have successively:
||u + v||2 = (u + v) (u + v) = ||u||2 + 2u v + ||v||2 = ||u||2 + ||v||2 .
A vector u = (u1 , . . . , un ) Rn can be naturally identified with a column or with a row matrix,
namely with

u=

u1
u2
..
.

or u = [u u un ].
1
2

un
Consequently, the set Rn of n-tuples can be naturally identified with either the space of column
matrices, or with the space of row matrices. The sum u + v of two vectors u = (u1 , . . . , un ), v =
(v1 , . . . , vn ) will be identified consquently

u
1

u2
u+v =
..
.

identified with the sum

v1
u + v1

1


v2 u2 + v2
+

.. =
..
.
.


vn
un + vn
un

45

of the column matrices u and v, or with the sum

u + v = [u1 u2 un ] + [v1 v2 vn ] = [u1 + v1 u2 + v2 un + vn ]
of the row matrices u and v.
Similarly, the scalar multiple ku of the real k with the ordered n-tuple u = (u1 , u2 , , un ) can
be either identified with the scalar multiple

ku = k

ku1

u1

ku2

=
.. ,
.

un
kun
u2
..
.

or with the scalar multiple

ku = k[u1 , u2 , . . . , un ] = [ku1 , ku2 . . . , kun ].

3.2

Linear transformations from Rn to Rm

Definition 3.2.1 A mapping TA : Rn Rm , TA (x) = Ax, where A is an m n matrix, is called

linear transformation, or linear operator, if m = n. The linear transformation TA is equally called
the multiplication by A. If A = [aij ]mn , then

x1
x2
..
.
xn

a11

a
= 21

..

am1

a12

...

a1n

a22
..
.

...
..
.

a2n
..
.

am2

. . . amn

x1
x2
..
.
xn

a11 x1 + a12 x2 + + a1n xn

a21 x1 + a22 x2 + + a2n xn
=

..

.

am1 x1 + am2 x2 + + amn xn

Observe that

1
0
..
.
0

a11

= 21
..

am1

,T

0
1
..
.
0

a12

= 22
..

am2

,...,T

0
0
..
.
1

a1n

= 2n
..

amn

The matrix A is called the standard matrix of the linear transformation T and it is usually
denoted by [T ] and the multiplication by A mapping is also denoted by TA . Therefore [T ] =

46

. .
.
[T (e1 )..T (e2 ).. ..T (en )], where

e1 =

1
0
..
.
0

1
0
, e = , . . . , en = .
2 ..
..

.
.

0
1

1. The orthogonal projection of R2 on the x-axis, px : R2 R2 , px (x1 , x2 ) = (x1 , 0) is a linear

mapping since

and [px ] =

x1

x2

x1

1 0

0 0

x1

x2

1 0

. Its equations are :

0 0

w1

= x1

w2

0.

2. The orthogonal projection of R2 on the y-axis, py : R2 R2 , py (y1 , y2 ) = (0, y 2 ) is a linear

mapping since

and [py ] =

y1

y2

y2

0 0

0 1

y1

y2

0 0

. Its equations are :

0 1

w1

w2

= y2 .

3. The reflection of R3 about the x1 x2 -plane, Sx1 x2 : R3 R3 , Sx1 x2 (x, y, z) = (x, y, z) is a

linear mapping since

y = y
x
1 2

z
z

Sx

1 0

= 0 1
0

0 0 1

y

z

1 0

and [Sx1 x2 ] = 0 1
0

0 0 1

w1

. Its equations are : w2

w3

= z

4. The reflection of R3 about the x1 x3 -plane, Sx1 x3 : R3 R3 , Sx1 x3 (x, y, z) = (x, y, z) is a

linear mapping since

0 0

= 0 1 0
y
y
1 x3

z
z
0
0 1

Sx

47

y

z

1
0 0

and [Sx1 x3 ] = 0 1 0

0
0 1

w1

. Its equations are : w2

w3

= y
=

5. The reflection of R3 about the x2 x3 -plane, Sx2 x3 : R3 R3 , Sx2 x3 (x, y, z) = (x, y, z) is a

linear mapping since

1 0 0

Sx x y = y = 0 1 0
2 3

z
z
0 0 1

w1 = x
1 0 0

and [Sx2 x3 ] = 0 1 0 . Its equations are : w2 =

y

0 0 1
w3 =
z

y

z

6. The orthogonal projection of R3 on the x1 x2 -plane, Px1 x2 : R3 R3 , Px1 x2 (x, y, z) = (x, y, 0)

is a linear mapping since

y = y =
1 x2


z
0

1 0 0
w1

and [Px1 x2 ] = 0 1 0 . Its equations are : w2

0 0 0
w3
Px

1 0 0

0 1 0 y

0 0 0
z
= x
= y
= 0

7. The orthogonal projection of R3 on the x1 x3 -plane, Px1 x3 : R3 R3 , Px1 x3 (x, y, z) = (x, 0, z)

is a linear mapping since

y = 0 =
x
1 3


z
z

1 0 0
w1

and [Px1 x3 ] = 0 0 0 . Its equations are : w2

0 0 1
w3
Px

1 0 0

0 0 0 y

0 0 1
z
= x
= 0
= z

8. The orthogonal projection of R3 on the x2 x3 -plane, Px2 x3 : R3 R3 , Px2 x3 (x, y, z) = (0, y, z)

is a linear mapping since

0 0 0

=
=

y
y
y
0
1
0
2 x3



z
z
0 0 1
z

Px

48

0 0 0
w1

and [Px2 x3 ] = 0 1 0 . Its equations are: w2

0 0 1
w3

= 0
= y
= z

9. The rotation operator of R2 through a fixed angle ,

R : R2 R2 , R = (x cos y sin , x sin + y cos )
is a linear mapping since

x
x cos y sin
cos sin
x
=

R =
x sin + y cos
sin
cos
y
y

and [R ] =

cos sin
sin

. Its equations are:

cos

w1

x cos y sin

w2

= x sin + y cos .

Indeed, the rotation operator R rotates the point (x, y) = (r cos , r sin ), counterclockwise
with the angle if > 0 and clockwise if < 0, the coordinates of the rotated point being
(w1 , w2 ) = (r cos( + ), r sin( + )), such that one gets

w1

x cos y sin

w2

= x sin + y cos .

10. The rotation operator of R3 through a fixed angle about an oriented axis, rotates about the
axis of rotation each point of R3 in such a way that its associated vector sweeps out some
portion of the cone determine by the vector itself an by a vector which gives the direction
and the orientation of the considered oriented axis. The angle of the rotattion is measured
at the base of the cone and it is measured clockwise or counterclockwise in relation with a
viewpoint along the axis looking toward the origin. As in R2 , the positives angles generates
counterclockwise roattions and negative angles generates clockwise roattions. The counterclockwise sense of rotaion can be determined by the right-hand rule: If the thumb of the right
hand points the direction of the direction of the oriented axis, then the cupped fingers points
in a counterclockwise direction. The rotation operators in R3 are linear. For example
an angle has
(a) The counterclockwise rotation about the positive x-axis through

the
w1 = x
1
0
0

equations w2 = y cos z sin , its standard matrix is 0 cos sin .

w3 = y sin + z cos
0 sin cos
an
(b) The counterclockwise rotation about the positive y-axis through

w1 = x cos + z sin
cos

equations w2 = y
, its standard matrix is 0

w3 = x sin + z cos
sin

49

angle has the

0 sin

.
1
0

0 cos

(c) The counterclockwise rotation about the positive z-axis through

an angle

w1 = x cos y sin
cos sin

equations w2 = x sin + y cos , its standard matrix is sin

cos

w3 = z
0
0

has
the
0

0 .

(d) The homotopy of ratio k R is the linear operator Hk : Rn Rn , Hk (x) = kx, Its
standard matrix is

k 0 0

0 k 0

.. .. . . .. and its equations are

. .
. .

0 0 k

w1

kx1

w2

=
..
.

kx2

wn

= kxn .

If 0 k 1, then Hk is called contraction, and for k 1, Hk is called dilatation.

Let A, B be matrices of sizes nk and k m respectively, and TA : Rn Rk , TB : Rk Rm
are the linear transformations of standard matrices A and B respectively, namely
T A (x) = Ax, T B (y) = By.
Then the composed mapping TB TA is also linear and TB TA = TBA . Indeed
(TB TA )(x) = TB (TA )(x)) = TB (Ax) = B(Ax) = (BA)x = TBA (x).
We have just proved that [T2 T1 ] = [T2 ][T1 ] for any two linear mappings whose composition
T2 T1 is defined. Similarly, if the composed map T3 T2 T1 , of the linear mappings T3 ,T2 ,T1 ,
is defined, then [T3 T2 T1 ] = [T3 ][T2 ][T1 ].
Let T1 : R2 R2 be the reflection operator about the line y = x and let T2 : R2 R2 be
the orthogonal projection about the y-axis. Find T1 T2 , T2 T1 and their standard matrices.
Show that R R = R R and it is the rotation of angle 1 + 2 .
1

3.3

Properties of linear transformations from Rn to Rm

Definition 3.3.1 A linear transformation T : Rn Rm is said to be one-to-one if u1 , u2

Rn , u1 6= u2 T u1 6= T u2 .
It follows that each vector w in the range {T x : x Rn } of an one-to-one linear transformation
T : Rn Rm there is exactly one vector x such that T x = w.
Theorem 3.3.2 If A is an n n matrix and TA : Rn Rn is the multiplication by A, then the
following statements are equivalent:

50

1. A is invertible.
2. The range of TA is Rn .
3. TA is one-to-one.
If TA : Rn Rn is a one-to-one linear operator, then the matrix A is invertible and TA1 : Rn Rn
is also linear. Moreover,
TA (TA1 (x)) = AA1 x = x = TI x = idRn (x), x Rn .
TA1 (TA (x)) = A1 Ax = x = TI x = idRn (x), x Rn .
Consequently TA is also invertible and its inverse is TA1 = TA1 . Therefore, for the standard matrix
of the inverse of an one-to-one linear operator T : Rn Rn , we have [T 1 ] = [T ]1 .
Examples 3.3.3 The rotation operator of R2 through a fixed angle ,
R : R2 R2 , R = (x cos y sin , x sin + y cos )

cos sin
, which is invertible
is a linear mapping since has the standard matric and [R ] =
sin
cos

cos sin
cos() sin()
=
= [R ]. Consequently R is invertible
and [R ]1 =
sin cos
sin()
cos()
and R1 = R .
Theorem 3.3.4 A transformation T : Rn Rm is linear if and only if
1. T (u + v) = T (u) + T (v) for all u, v Rn .
2. T (cu) = cT (u) for all u Rn .

Proof. Indeed, if T = TA , then T = TA , namely TA (u + v) = A(u + v) = Au + Av = TA u + TA v.

Conversely, if the given conditions are satisfied, then one can easily show that for any k vectors
u1 , . . . , uk Rn we have T(u1 + + uk ) = T(u1 ) + + T(uk ). Oncan now show that T = TA ,
. .
where A = [T(e ).. ..T(e )], where
1

e1 =

1
0
..
.
0

,e =
2

0
1
..
.
0

51

,...,e =
n

0
0
..
.
1

Indeed, for

x=

x1
x2
..
.
xn

=x
1

1
0
..
.
0

+x
2

0
1
..
.

+ + xn

0
0
..
.

= x e + x e + + xn e
2 2
1 1
n

we have
T(x) = T(x1 e1 + x2 e2 + + xn en ) = x1 T(e1 ) + x2 T(e2 ) + + xn T(en ) = Ax.

52